digital-human-koubo
AI digital-human talking-head videos from one photo: Renoise MCP generation + 5 editing styles + 48 creator-grade packaging components. Agent skill for Claude Code / Codex / OpenClaw / WorkBuddy …
- Stars
- 686
- Forks
- 182
- + today
- +2
- Created
- 1d
Ranking data as of October 26, 2026 (UTC).
Star History
Today, hour by hour
Overview
AI digital-human talking-head videos from one photo: Renoise MCP generation + 5 editing styles + 48 creator-grade packaging components. Agent skill for Claude Code / Codex / OpenClaw / WorkBuddy … It ranks #135 on GitTiger, gaining +2 stars on October 26, 2026 (UTC).
The project is written in TypeScript and has 182 forks. It was created 1d ago.
What it is
`andyhuo520/digital-human-koubo` packages a workflow for making presenter-led talking-head clips with a synthetic person. Its description sums the idea up as turning a single photograph into video, and the code is TypeScript under an MIT licence. The same description calls it an agent skill for assistants such as Claude Code, Codex, OpenClaw and WorkBuddy.
How the pipeline is described
- **A persistent presenter.** Face, wardrobe, voice notes and a prompt template sit in `persona.json`, so each video shows the same on-screen person. - **Scene compositing.** GPT Image places the presenter into a chosen setting first; the preview lists a studio, shop, classroom and street. - **Speech-driven footage.** Narration is split into chunks of fifteen seconds or less, then turned into lip-synced video with Grok Video 1.5 called through Renoise MCP. - **Named alternatives.** The preview also mentions Wan 3.0 and OmniHuman, without explaining how either is wired in. - **Packaging.** The description claims five editing styles plus forty-eight reusable presentation components.
Getting started
The visible README text ends before any setup material, so dependencies, service credentials and runtime requirements are absent. Consult the repository README for the real procedure.
Caveats
Style and component counts are the description's promotional wording, not verified behaviour, and nothing supplied here covers output quality or asset licensing.
git clone https://github.com/andyhuo520/digital-human-koubo.git
cd digital-human-koubo
# see README for setupFrequently Asked Questions
Does each video use a different synthetic presenter?
No. One presenter description is stored in persona.json and reused, so every clip in a series keeps the same face, wardrobe and voice character.
Which generation services does the workflow rely on?
The README preview names GPT Image for building the scene and Grok Video 1.5, reached through Renoise MCP, for the lip-synced talking segments. Wan 3.0 and OmniHuman are also mentioned without detail.
How long are the generated talking segments?
The script is divided so that no single segment runs longer than fifteen seconds, and each chunk becomes its own lip-synced clip.